Building systems that decide whether AI-generated actions are allowed to execute.
AI systems can propose actions.
But what actually stops a bad one from executing?
Most systems:
- detect issues
- log them
- audit after the fact
I focus on something else:
stopping invalid actions before they execute.
A working end-to-end pipeline that:
- takes a proposed action
- evaluates it against governance rules
- decides:
- ✅ COMMIT ALLOWED
- ❌ COMMIT BLOCKED
🔗 https://github.com/Waveframe-Labs/governed-finance-mutation-demo
Policy
↓
Contract Compiler
↓
Proposal
↓
Normalizer
↓
CRI-COR
↓
COMMIT ALLOWED / BLOCKED
Deterministic enforcement engine that decides whether a system state change is allowed to commit.
🔗 https://github.com/Waveframe-Labs/CRI-CORE
Turns governance rules into executable contracts.
🔗 https://github.com/Waveframe-Labs/cricore-contract-compiler
Converts actions into a standard structure for enforcement.
🔗 https://github.com/Waveframe-Labs/proposal-normalizer
This work is developed under Waveframe Labs, where I’m building infrastructure for governed, auditable AI systems.
The focus is the execution boundary: the moment a system attempts to act.
📧 swright@waveframelabs.org
🌐 https://waveframelabs.org
🧭 https://orcid.org/0009-0006-6043-9295


